AI: Hope for Developing Economies, Risk for Developed

The United States and other developed nations exhibit significant skepticism toward AI's future impact.

In contrast, poorer, developing countries express substantially greater optimism regarding AI's potential for economic advancement.

While traditional economic analyses often predict AI will exacerbate global inequality due to existing infrastructure gaps, developing nations perceive AI as a direct solution to critical shortages in areas such as food, health, and education.

Developing nations experience acute shortages of essential personnel, such as nurses and educated teachers, leading to critical service gaps.

Unlike in the United States, where AI's application often involves automating existing jobs, developing countries deploy AI to address these human capital deficits.

The automation of specific tasks can significantly enhance the productivity of the limited personnel available in these regions.

The core issue in developing economies is a scarcity of human capital, not a surplus that AI might displace.

Companies in developing nations frequently benefit from protection or maintain close ties with government leaders, limiting their incentive for disruptive innovation.

Substantial changes in firm operations, even those driven by efficiency gains from AI, may require prior shifts in governmental policies or leadership.

Experience from competition policy demonstrates that market entry and the adoption of new technologies are frequently obstructed by structural factors that go beyond simple economic efficiency.

AI functions as a multifaceted technology that not only offers direct applications but also spawns and enhances other general-purpose technologies.

Software coding tools, a secondary layer enabled by AI, provide an accessible entry point for digitization in developing economies.

By reducing the cost of coding tools, AI facilitates affordable and cost-effective digitization efforts in these regions.

The most significant impact of AI lies in its ability to enable the creation of applications tailored to function effectively within low-resource environments.

Sub-Saharan Africa has only 2.9 physicians per 10,000 people, significantly below the global average of 16, highlighting a severe healthcare deficit.

Low-income countries contend with an average of 40 students per teacher, starkly contrasting with rich countries' average of 15 students per teacher.

Teachers in low-income settings frequently lack access to digital resources essential for effective lesson planning and instructional support.

Farmers in these regions often lose over 50% of their potential profits to middlemen, primarily due to a lack of transparency in market prices.

GitHub data indicates substantial AI-related development activity in countries such as India, Pakistan, Bangladesh, and Nigeria.

These regions currently serve as major hubs for multinational outsourcing operations, leveraging their skilled labor forces.

Multinational corporations possess the flexibility to swiftly relocate operations between countries, driven by AI-powered cost optimization strategies.

This agility poses an immediate risk that local outsourcing jobs could rapidly disappear if AI leads to consolidation or automation within these global operations.

Initial academic research on AI predominantly concentrated on labor economics, examining issues of inequality and potential job displacement.

Christopher Tonetti is pursuing an empirically grounded understanding of AI's impact on aggregate economic growth, moving beyond individual labor market effects.

His model utilizes historical data, including past automation trends in software development, to forecast future economic output and growth patterns.

When a specific task experiences a significant surge in productivity, the compensation for that task becomes directly dependent on its scarcity within the economy.

If automation makes a task abundant, its economic value as a driver of returns diminishes substantially.

Total economic output is perpetually limited by the input that remains in the smallest quantity, highlighting the critical role of bottlenecks.

In an economy characterized by weak links, abundant resources yield minimal returns, while scarce resources attract the majority of capital flow.

Early in the diffusion of computers, the cost share dedicated to these technologies rose significantly.

However, recent decades have shown a marked decline in spending on computers, despite their continued productivity enhancements.

This trend parallels historical shifts in sectors like agriculture, where high productivity eventually leads to a lower proportion of income being spent on the good itself.

Approximately 2% of human-performed tasks are being replaced by machines each year.

A Sigma value of 0.2 indicates an economic environment characterized by 'weak links,' where overall productivity is constrained by the least productive tasks.

The economy is frequently bottlenecked by tasks that exhibit the lowest levels of productivity, limiting aggregate growth.

Under a 'business-as-usual' scenario, machine productivity is projected to grow approximately 4% faster than human productivity on an annual basis.

A more aggressive scenario, termed 'Moore's Law everywhere,' posits that machine productivity could grow 11 percentage points per year faster than human productivity, mirroring advancements in computing and IT.

This modeling approach helps distinguish between long-run theoretical discussions about AI's impact and its more immediate effects on economic growth.

Such scenarios provide frameworks for understanding potential future economic trajectories by drawing comparisons to past technological acceleration.

Aggregate economic growth is fundamentally determined by the 'weak links' in a system, rather than solely by advancements at the technological frontier.

The widespread adoption of AI that performs numerous tasks at a medium level of quality is more beneficial for overall growth than AI achieving highly specialized breakthroughs in very narrow applications.

Historically, economic growth in the U.S. has been driven more by expanding the range of tasks performed by machines rather than by focusing exclusively on human task performance.

Therefore, fostering broad-based AI capabilities across many sectors can lead to faster and more substantial aggregate growth than hyper-specialized AI advancements.

High potential returns from innovation do not automatically guarantee rapid adoption; historical data indicates that significant innovations can take decades to fully diffuse across an economy.

Companies, such as Citibank, possess tools beyond mere efficiency, including lobbying efforts and regulatory protection, which can be deployed to resist disruptive technological changes.

Agency frictions arise when workers resist adopting new technology due to concerns about job loss or income reduction, making it challenging for management to implement changes.

Mid-sized companies are increasingly hesitant to reorganize their production processes around current AI technologies because of the exceptionally rapid pace of AI development and change.

This leads to a 'wait-and-see' investment strategy, where firms defer significant capital expenditures, anticipating that other companies will develop and stabilize AI solutions.

The rationale behind this approach is to avoid making obsolete capital investments by allowing the market to mature and then selecting the most effective and proven AI choices.

An increase in tasks performed by machines does not inherently lead to a reduction in the income share of human labor.

For the standard of living, the overall level of total economic output is more critical than the specific share of that income allocated to labor.

Economic welfare can significantly improve if total income experiences substantial growth, even if there is a concurrent shift resulting in labor receiving a smaller share of that increased income.

A history of systemic failures in cybersecurity makes public distrust in institutions' ability to solve these issues rational and understandable.

An excessive focus on AI risks related to job displacement, which may be less pervasive, distracts from more immediate and prevalent short-run threats, particularly in cybersecurity.

Economic research should guide policymakers in allocating resources towards the most productive and essential security directions, addressing actual and present dangers effectively.